Despite persistent regularities in migration age patterns, data noise arises when migration data are
disaggregated by single-year age groups or derived from small-sized samples, requiring observed data
to be smoothed. This paper proposes a one-step non-parametric kernel modeling approach to smooth
age-specific migration data, and seeks to evaluate its performance against conventional smoothing
techniques, both parametric (model schedules) and non-parametric (cubic splines). Model schedules
impose a proto-typical age profile of migration, and both model schedules and cubic splines require
iterative fitting. Drawing on the age profiles of 28 countries, the paper establishes through residual
analysis that one-step kernel regressions better captures the underlying structure of migration than
conventional iterative smoothing methods. Application of kernel regressions to migration in Chile
since 1970 reveals significant shifts in the age-structure of migration concealed by model schedules.